Patent Application
20110187399
Embodiments according to the invention are related to a signal distribution structure and to a method for distributing a signal from a driver to a plurality of devices.
Some embodiments according to the invention are related to a concept for a 50Ω by by-4 Y-sharing and by-2 sharing,
Some embodiments according to the invention can be used as a solution for massively parallel high-speed DRAM tests.
In many applications, it is desirable to distribute a signal from a signal source to a plurality of signal sinks. For example, a distribution of a signal from a signal source to a plurality of signal sinks is useful, whenever a plurality of devices or components is to be supplied with identical input signals. However, signal integrity is often an issue in such applications.
According to an embodiment, a signal distribution structure for distributing a signal to a plurality of device connections may have: a first signal guiding structure including a first characteristic impedance; a node, wherein the first signal guiding structure is coupled to the node; a second signal guiding structure including one or more transmission lines, wherein the one or more transmission lines of the second signal guiding structure are coupled between the node and the plurality of device connections, and wherein the second signal guiding structure includes, side-viewed from the node, a second characteristic impedance which is lower than the first characteristic impedance; and a matching element connected to the node; wherein the matching element is configured to match an impedance at the node, side-viewed from the second signal guiding structure, to the second impedance, while increasing a mismatch between an impedance at the node, side-viewed from the first signal guiding structure, and the first impedance.
According to another embodiment, a method for distributing a signal from a driver to a plurality of devices may have the steps of: providing a signal to a node via a first signal guiding structure including a first characteristic impedance; forwarding a portion of the signal incident to the node via the first signal guiding structure to the plurality of devices, wherein said portion of the signal is forwarded to the devices via a second signal guiding structure; reflecting another portion of the signal incident to the node via the first signal guiding structure back to the first signal guiding structure; and forwarding a signal portion of a signal incident to the node via the second signal guiding structure to the first signal guiding structure and to the matching element, while suppressing a reflection of said signal, incident to the node via the second signal guiding structure, back to the second signal guiding structure.
As an example only, out of a large variety of possible applications, requirements in solutions from the field of device testing will be described in the following.
In some applications, so-called “driver sharing” is used. Regarding the concept of “driver-sharing”, it should be noted that a traditional test interface for automated test equipment (ATE) may, for example, use a point-to-point connection between tester resources (for example tester output channels and/or tester input channels), and a device-under-test (DUT). However, for cost sensitive applications, a plurality of devices, for example between two and thirty-two, or 64, or 128, or 256, or 512, . . . devices under test, may be tested in parallel. However, the testing of some devices such as, for example, DRAM testing, may entail a massive parallel testing to achieve the cost-of-test goals.
In some cases, in production a parallelism of a minimum of 64 devices under test may be useful. In other words, it may sometimes be desirable to test 64 devices, or even more devices, using a single tester. An economical way of achieving this may comprise sharing tester resources among devices-under-test. Since, for example, for dynamic random access memories (DRAMs), the number of inputs is-in some cases much higher than the number of outputs, a sharing of driver channels of the automated test equipment (ATE) may be particularly attractive.
However, in some cases, a reduced signal quality may be taken into account as a compromise, when drivers are shared. In particular, a reduction of the signal quality may occur at high speed.
In the following, the concept of shared drivers and non-shared drivers will be briefly explained taking reference to
a shows a block schematic diagram of a device-under-test interface for a traditional parallel testing. In contrast,
The test arrangement shown in
As can be seen from
However, now taking reference to
To summarize the above, the concept of shared drivers versus non-shared drivers has been schematically described with reference to
In the following, a plurality of conventional sharing concepts will be described taking reference to
Conventionally, two topology schemes are often used for driver sharing. For example, a so-called “Y-sharing”, which is also designated as “fork” or “fork sharing”, may be used. Alternatively, a so-called “Daisy-Chain”, which may also be designated as “multidrop bus”, “tapped bus”, or “fly-by”, may be used. Taking reference to
Moreover, a first device under test 840 (or an input thereof, or an input/output thereof) may be coupled to the second transmission line 820, as shown in
It should be noted here that a matching condition is obtained at the node 830 for signals or waves traveling in both directions. Signals incident to the node 830 from the first transmission line 812 will “see” an impedance of 50Ω, as the “joint” characteristic impedance of the second transmission line 820 and third transmission line 822, side-viewed from the node 830, is 50Ω. Signals (or waves) which are reflected by the devices under test 840, 842, and which come back from the devices under test, do not find a matched impedance, but find an Impedance of 50Ω in parallel with 100Ω (50Ω∥100Ω). The two reflections cancel out each other. This can be noticed if, for example, a 50Ω termination is applied at the position of the second device under test 842 in order to prevent a reflection at this position. In this case, the reflections do no longer cancel out at the node 830, and massive distortions appear. The main operating principle is an opposite erasement or cancellation of reflections.
Thus, there will be no reflection (or only a negligible reflection) at the node 830, if signals are reflected by the devices under test 840, 842. Thus, signals reflected at the devices under test 840, 842 will travel back to the buffer or driver 810 via the first transmission line 812, and may be absorbed in the driver 810. However, the price for this matching condition is the need to fabricate transmission lines having a comparatively high impedance of 100Ω, which is challenging in some transmission line fabrication technologies.
In the following, the so-called “Daisy-Chain” topology will be described taking reference to
In the following, some problems arising from the abovementioned topologies (Y-sharing topology and Daisy-Chain topology) will be discussed. It will be assumed that said conventional topologies would be used for a BY-4 sharing. It should be noted that in the following, only a single driver is shown, while the concept can naturally be extended to test arrangements comprising more than one driver.
To summarize the above, using a Y-sharing topology for implementing a BY-4 sharing brings along the difficulty that transmission lines comprising comparatively high characteristic impedance need to be fabricated. However, the fabrication of transmission lines comprising comparatively high characteristic impedance is sometimes difficult and/or costly.
In the following, details regarding the Daisy-Chain topology will be described.
In the following, disadvantages of the Daisy-Chain concept will subsequently be described taking reference to
As indicated at reference numeral 1150, each tap of the tapped transmission line 1020 may cause a reflection. The reflection may for example originate from the stub branching from the tapped transmission line 1020, and also from the parasitic input capacitances 1130a to 1130d of the devices 1030a to 1030d.
The reflections caused by the taps of the tapped transmission line 1120, and by the input of the devices under 1030a to 1030d may result in a degradation of the signals, as shown at reference numeral 1170.
A signal representation at reference numeral 1170 describes the signal seen at the input of the first device under test 1030a. An abscissa 1172 describes a time, and an ordinate 1174 describes a signal at the input of the first device under test 1030a. As can be seen from the graphical representation at reference numeral 1170, the signal at the input of the first device under test 1030a, which is represented by a line 1176, is distorted by reflections 1178a, 1178b, and 1173c from the second device under test, from the third device under test and from the fourth device under test. The distortion caused by the reflections is the stronger the steeper the signal transition of the signal generated by 1010 are. To summarize the above,
In the following, some pro's (or advantages) and con's (disadvantages) for the two topologies mentioned above will briefly be discussed.
Y-Sharing:
Daisy-Chain:
In view of the above, there is a need for a concept of forwarding a signal to a plurality of devices, which brings along a good compromise with respect to signal integrity and production costs.
Some embodiments according to the invention create a signal distribution structure for distributing a signal to a plurality of devices. The signal distribution structure may comprise a first signal guiding structure, comprising a first characteristic impedance. The signal distribution structure may further comprise a node, wherein the first signal guiding structure is coupled to the node. The signal distribution structure may also comprise a second signal guiding structure comprising one or more transmission lines. The one or more transmission lines of the second signal guiding structure may be coupled between the node and a plurality of device connections. The second signal guiding structure may comprise, side-viewed from the node, a second characteristic impedance which is lower than the first characteristic impedance. The signal distribution structure may comprise a matching element connected to the node. The matching element may be configured to match an impedance at the node, side-viewed from the second signal guiding structure, to the second impedance, while increasing the mismatch between an impedance at the node, side-viewed from the first signal guiding structure, and the first impedance.
For example, assuming that the impedance of the first signal guiding structure is higher than the impedance of the second signal guiding structure, a matching between the first signal guiding structure and the second signal guiding structure in the absence of the matching element can be characterized by a reflection coefficient. In the absence of the matching element, a magnitude of the reflection coefficient may be determined by the characteristic impedances of the first signal guiding structure and the second signal guiding structure.
However, in the presence of the matching element, a first reflection coefficient, describing a reflection of a wave incident via the first signal guiding structure, may be determined by the characteristic impedance of the first signal guiding structure and an impedance of a parallel circuit of the second signal guiding structure and the matching element. The impedance of said parallel circuit may be lower than the characteristic impedance of the second signal guiding structure. Accordingly, a mismatch for waves incident via the first signal guiding structure is increased.
Also, in the presence of the matching element; a second reflection coefficient, describing a reflection of a wave incident via the second signal guiding structure, may be determined by the characteristic impedance of the second signal guiding structure and by an impedance of a parallel circuit of the first signal guiding structure and the matching element. The impedance of said parallel circuit may approximate the characteristic impedance of the second signal guiding structure. Accordingly, a mismatch for waves incident via the second signal guiding structure may be reduced in the presence of the matching element, when compared to a case in the absence of the matching element. Some embodiments according to the invention are based on the finding that a signal transmission or a signal distribution from the first signal guiding structure to the devices connected to the second signal guiding Structure can be performed with good signal integrity and at reasonable cost, if an impedance mismatch of signals traveling towards the node via the first signal guiding structure is tolerated. However, it has also at the same time been found that the signal integrity can be significantly improved, if an impedance match is achieved, for signals reflected from the devices, which reflected signals are traveling towards the node via the second signal guiding structure. Thus, while costs can be reduced by allowing a mismatch in a forward signal transmission direction (i.e. from the first signal guiding structure towards the second signal guiding structure), signal integrity can be ensured by providing matching in a backward signal transmission direction (i.e. from the second signal guiding structure towards the first signal guiding structure).
Also, if the second signal guiding structure comprises multiple conductors coupled to the node, reflections traveling towards the node via the multiple conductors may cancel out, at least partially, due to the presence of the matching element. For example, if the second signal guiding structure comprises two conductors, waves traveling towards the node concurrently via the two conductors may be reflected at the node, but the reflections may cancel out at least partially.
It has been found that coupling a matching element to the node may be used to provide the matching in the backward signal transmission direction, if the characteristic impedance of the second signal guiding structure is lower than an impedance of the first signal guiding structure. However, it has also been found that an increase of a mismatch in the forward signal transmission direction, which is caused by the matching element, is tolerable in many environments and does not severely degrade the signal integrity. In other words, it has surprisingly been found that the advantage resulting from an improvement of the matching in the backward signal transmission direction, which improvement is caused by the presence of the matching element, strongly outweighs the disadvantages caused by a deterioration of the matching in the forward signal transmission direction, which deterioration is also caused by the matching elements.
Embodiments of the present invention will be detailed subsequently referring to the appended drawings, in which:
a and 2b show block schematic diagrams of signal distribution structures according to an embodiment of the invention;
a and 3b show block schematic diagrams of signal distribution structures according to embodiments of the present invention;
a, 4b, and 4c show a graphical representation of matching conditions;
a shows a block schematic diagram of a device-under-test interface for traditional parallel testing;
b shows a block schematic diagram of a driver sharing device-under-test interface for massive parallel testing;
a shows a block schematic diagram of a conventional Y-sharing topology;
b shows a block schematic diagram of a conventional Daisy-Chain topology;
In the following, different embodiments according to the invention will be explained taking reference to
Moreover, the signal distribution structure 100 comprises a matching element 140 connected to the node. The matching element 140 is configured to match an impedance ZSV2 at the node, side-viewed from the second signal guiding structure 130, to the second impedance (impedance or overall impedance ZTL2 of the second signal guiding structure). The matching element 140 also increases a mismatch, for example as explained above, between an impedance ZSV1 at the node, side-viewed by the first signal guiding structure 110, and the first impedance ZTL1 (the impedance of the first signal guiding structure 110.
Moreover, it should be noted that the second signal guiding structure is typically coupled to a plurality of device connections 132a to 132d.
In the following, the functionality of the signal distribution structure 100 will be described. It is assumed here that it is desired to distribute a signal from a first end 112 of the first signal guiding structure 110 towards the device connections 132a to 132d via the first signal guiding structure 110, the node 120, and optionally the second signal guiding structure 130. A signal which is fed to the first end of the first signal guiding structure may propagate via the first signal guiding structure 110 towards the node. As the impedance ZSV1 at the node, side-viewed from the first signal guiding structure 110, is mismatched with respect to the impedance ZTL1 of the first signal guiding structure, a portion of the signal energy is reflected back into the first signal guiding structure 110. Another portion of the signal energy is dissipated in the matching element 140. However, yet another portion of the signal energy propagates towards the device connections 132a to 132d via the second signal guiding structure 130 which in some embodiments may have zero length (vanishes).
To summarize the above, if a signal is fed to the first end 112 of the first signal guiding structure 110, a portion of said signal is forwarded to the device connections 132a to 132d, and another portion of the signal is reflected back to the first end 112 of the first signal guiding structure 110. However, assuming that the first end 112 of the first signal guiding structure is terminated with an impedance approximating the characteristic impedance ZTL1 of the first signal guiding structure (or being the complex-conjugate thereof in an ideal case), multiple reflections can be avoided. Thus, as an effect, multiple reflections when forwarding a signal from the first end 112 of the first signal guiding structure 110 towards the device connections 132a to 132d can be avoided.
In the following, it is assumed that a portion of a signal provided to the device connections 132a to 132d is reflected, for example, because the inputs of the devices connected to the device connections 132a to 132d are mismatched with respect to the second signal guiding structure 130.
For example, the connections between the second signal guiding structure 130 and the device connections 132a to 132d may comprise transmission lines T13a to T13d, each having a characteristic impedance Zt13. The reflection at a device connected to one of the device connections 132a to 132d is determined by the fact that the device impedance (or device input impedance) does not match the characteristic impedance Zt13. In many cases, the device impedance is a high impedance, or is a capacitive impedance. Thus the signal is reflected back into the transmission lines T13a to T13d at the device connections 132a to 132d. As these reflections occur with the same phase at all four devices (assuming the devices are sufficiently similar), the reflections add up at the node 125, where all four transmission lines T13a to T13d converge. Accordingly, there is only a signal traveling back to towards the node 120, but there are no signals (or only negligible signals) traveling back towards the device connections 132a to 132d. Zt13 may be chosen such that it fits Zt12. For example, in a by-4-sharing, the relationship Zt13=4*Zt12 may be fulfilled.
A Signal reflected back from the node 125 may propagate towards the node 120 via the second signal guiding structure 130. However, as was previously discussed, the impedance at the node 120, side-viewed from the second signal guiding structure 130 (which impedance is designated with ZSV2) is matched to the characteristic impedance ZTL2 of the second signal guiding structure. Thus, signals reflected by the devices and propagating towards the node 120 via the second signal guiding structure 130 will not be reflected back towards the devices when arriving at the node 120, due to the impedance at the node, side-viewed from the second signal guiding structure, being matched to the impedance of the second signal guiding structure. Accordingly, the signals reflected back from the devices will not result in a multiple reflection, which could lead to severe signal degradation. Rather, a portion of the signals reflected by the devices will be dissipated in the matching element 140. Another portion of the signals reflected by the devices will propagate from the node 120 towards the first end 112 of the first signal guiding structure 110. Accordingly, if the first end 112 of the first signal guiding structure is possibly terminated, multiple reflections can be avoided.
To summarize the above, signal integrity may be maintained by providing a matching at the node 120 and node 125 for signals reflected back from the device connections 132a to 132d. However, allowing for a mismatch for signals propagating from the first end 112 of the first signal guiding structure 110 towards the device connections 132a to 132d, allows the use of the second signal guiding structure 130, an impedance of which is lower than an impedance of the first signal guiding structure 110 and a third impedance of T13a-d which is 50 ohm. Both can be fabricated easily in a standard PCB process. Accordingly, the cost efficiency can be improved by avoiding a need for fabricating high impedance signal guiding structures.
In the following, some possible implementations will be described taking reference to
a shows a block schematic diagram of a signal distribution structure, according to an embodiment according to the invention. The signal distribution structure shown in
Moreover, a matching element, for example a resistor 240 having a resistance RM may be coupled to the node 214. While a first terminal of the resistor 240 may be connected to the node 214, a second terminal of the resistor 240 may be coupled to a voltage source 242.
If N branch transmission lines are connected at the node 222, the equations
Zt12=Zt13/N
and
Rm=(Zt12*Zt11)/(Zt11−Zt12)
may hold (at least approximately).
In an advantageous embodiment, Zt13 and Zt11 may lie between 50 Ohm and 70 Ohm, because printed circuit board manufacturers can fabricate such transmission lines well, and because Zt12 becomes smaller in this case, which can also be fabricated well.
Regarding the functionality of the signal distribution structure 200, it should be noted that a signal can be forwarded from the connection 212 to the device connections 232a to 232d, or to the devices 234a to 234d.
In an embodiment, the following relationships may hold for the characteristic impedance ZTL1 of the first transmission line 210, for the characteristic impedance ZTL2 of the second transmission line 220, for the characteristic impedance ZTL3 of the branch transmission lines 232a to 232d, and for the impedance RM of the resistor 240;
Z
TL3
=Z
TL3
/N;
ZTL3=ZTL1; and
ZTL1//RM=ZTL2.
However, generally, Zt13 can be chosen freely in a range 0<Zt13<Zt11*N. Also, the equation
Rm=(Zt12*Zt11)/(Zt11−Zt12) may be fulfilled.
In some embodiments, impedances of 70 Ohm or 100 Ohm may be used for ZT13.
In the above equations, N designates the number of branch transmission lines 230a to 230d branching from the branching node 222. Naturally, some tolerances may occur. It has been found that a deviation from the above defined values by 30% (or even more) is well acceptable. However, if the deviations from the above defined values is less than 10%, a particularly good suppression of reflections can be achieved. The length of Zt12 (or of the transmission line 220) may be set to 0 with the effect that it can be omitted.
Taking into consideration the above described impedance values, the impedance situation as described with reference to
In an embodiment in which the lengths l1, l2, l3, l4 of the branch transmission lines 230a to 230d are, at least approximately identical, a matching condition is also fulfilled at the branching node 222 for signals reflected back from the device connections 232a to 230d. For example, it may be sufficient if the length of the branch transmission lines 232a to 232d do not differ by more than 10%. An even better matching can be achieved if the lengths do not differ by more than 5%.
In some embodiments, the connection 212, the transmission lines 210, 220, 230a to 230d and the device connections 232a to 232d may be arranged on a device-under-test board for usage in a device tester. The resistor 240 may also be placed on or in the device-under-test board. Thus, the signal distribution structure 200 may be used for distributing signals to a plurality of devices-under-test, when performing a device test.
Taking reference now to
The signal distribution structure 250 shown in
However, apart from the fact that the transmission line 220 of the signal distribution structure 200 is omitted, the electrical functionality of the signal distribution structure 250 is very similar to the functionality of the signal distribution structure 200. It should be noted here that the transmission lines 220a to 220d present a joint impedance to the node 214, which is determined by the parallel connection of the transmission lines 220a to 220d. Assuming that there are N transmission lines 220a to 220d having approximately identical impedances ZTL2, a joint impedance Zjoint of the transmission lines 220a to 220d is identical to ZTL2/N. It should be noted here that the transmission lines 220a to 220d can be considered as a second signal guiding structure, and that the joint impedance Zjoint can be considered as the impedance of the second signal guiding structure, side-viewed from the node 214.
Again, the first transmission line 210, the transmission lines 220a to 220d, the DUT connections 230a to 230d, and the resistor 240 may be arranged on (or in) a device-under-test board, for example for usage in combination with a device tester.
It should be noted that the branch point 214 may be implemented as a via. In some embodiments, the via forming the branch point 214 may be designed for good symmetry. Otherwise, some signal distortions may occur.
In the following, some modifications of the signal distribution structures 200, 250 will be described taking reference to
However, it should be noted that the connection 212 may be considered as a part of a first signal guiding structure. Nevertheless, an impedance of the first signal guiding structure, comprising the connection 212 and the first transmission line 210, is typically dominated by a characteristic impedance of the first transmission line 210, as the connection 212 is typically designed such that it forms a negligible impedance discontinuity.
Moreover, the signal distribution structure 300 may comprise a driver or buffer 320. An output of the driver or buffer 320 may be coupled to the first transmission line 210. Thus, the signal provided by the driver or buffer 320 can be forwarded to the devices 234a to 234d via the first transmission line 210, the node 214, the second transmission line 220, and the branch transmission lines 230a to 230d. In some embodiments, a signal degradation can be reduced by providing the driver 320 with an output impedance which is impedance matched with the characteristic impedance of the first transmission line 210. Thus, even if signals reflected back by the inputs of the devices 234a to 234d propagate to the output of the driver 320, the reflected signals are absorbed in the output impedance of the driver 320.
In some embodiments, the connection via 212a, the second transmission line 220, the branch transmission lines 230a to 230d and the device connections 232a to 232d may be arranged on (or in) a device-under-test board for usage in a device tester. Moreover, the resistor 240 may be arranged on (or in) the device-under-test board. In contrast, the driver 320, the first transmission line 210, and the connection pin 212b may for example be part of a device tester.
Taking reference now to
The connection via 300a, the branch transmission lines 320a to 320d and the device connections 323a to 323d may also be arranged on (or in) a device-under-test board, as explained above. In addition, the resistor 240 may be arranged on (or in) the device-under-test board. In contrast, the driver 320, the first transmission line 210, the connection pin 212b, and the voltage source or power supply 242 may be part of a device tester.
To summarize the above, a plurality of different possible arrangements has been described with reference to
In the following, the concept of impedance matching will briefly be explained taking reference to
Taking reference to
Taking reference now to
However, taking reference now to
However, it should be noted that the values shown here are to be considered as examples only. It should also be noted that in a real environment, a reflection coefficient for a wave traveling towards the node 430 via the second transmission line 420 can not normally be reduced down to zero. However, in some embodiments the reflection factor of such a wave traveling towards the node 430 can be reduced such that a magnitude of the reflection factor ρ is smaller than 0.3, or even smaller than 0.1.
It can also generally be said that the resistor 424 is configured to match the impedance at the node, side-viewed from the second transmission line 420, to the second impedance, i.e. to reduce a magnitude of the reflection factor ρ when compared to a case in which the resistor 424 is not present. In contrast, the presence of the resistor 424 typically increases the magnitude of the reflection coefficient ρ for a wave traveling towards the node 430 via the first transmission length 410, as shown in
Taking reference now to
The device-under-test board 530 may for example comprise a second transmission line 560, which second transmission line may for example comprise a characteristic impedance of Z=12.5Ω. An end of the second transmission line 560 may for example be coupled to a branching node 570. A plurality of branch transmission lines 580a to 580d may connect the branching node 570 with the device-under-test connections 582a to 582d of a plurality of devices under test 584a to 584d. In an embodiment, one branch transmission line 580a to 580b may be provided per device under test 584a to 584d. However, in some other embodiments a plurality of devices under test may be provided with input signals via a single one of the branch transmission lines 580a to 580d. A length of the second transmission line 560 may be zero. In other words, the second transmission line 560 can be omitted.
To summarize the above, in the embodiment shown in
To further summarize, some embodiments according to the invention are capable of avoiding at least some of the above-described disadvantages of the conventional Y-sharing topology, while keeping the key advantages.
In some embodiments according to the invention, one or more of the following effects can be obtained:
In some embodiments in accordance with the invention, the following compromise may occur:
In some embodiments, the devices-under-test may be chips comprising a terminal for a reference voltage Vref. In such an embodiment, a terminal of the resistor 554, which is opposite to the node 550, may be connected to said reference voltage. The reference voltage supplied to the devices-under-test may for example be used by the devices-under-test to distinguish the different logic levels. In other words, the reference voltage may for example be used by the devices-under-test to determine the threshold level for discerning between the different logic levels. Thus, by applying said reference voltage Vref to one terminal of the resistor 554, the signal transmission path (comprising the cables 520, the connection 540, and the transmission lines 560, 580a to 580d) may be biased in an efficient manner, such that reliable input levels can be applied to the inputs of the devices-under-test 584a to 584d, in spite of an attenuation effect caused by the matching concept described herein.
In some embodiments according to the invention, all Y-sharing branches may branch off from one point (which may also be designated as a branch point) with 50Ω impedance traces. In some embodiments, to match the joint impedance of the branches (all of the Y-sharing branches 580a to 580d), the sourcing trace (for example the transmission line 560) may have ¼ th of the branch impedance.
In some embodiments, to achieve back matching, the resistance in parallel to the driver cable impedance (for example an impedance of a parallel circuit of the cable 520 and the resistor 554) may have (at least approximately) the same impedance as the joint impedance of the branches (which may be equal to th of an individual branch impedance).
To summarize some of the aspects according to the present invention, a Y-sharing socketboard printed circuit board can be made “manufacturable” for higher sharing degree using the concept in accordance with the invention. For example, a Y-sharing socketboard may be designed for a by-4 sharing.
Simultaneously, the Y-sharing socketboard becomes suitable for high speed, due to the lower 50Ω branch impedance.
When a high symmetry is achieved (for example in the case of a low device-under-test input capacitance variation and in the case of a matched trace length), a significant increase in speed may be expected due to smaller reflections when compared to a Daisy-Chain topology.
According to some embodiments, the solution may fit the DDR3 and DDR4 minimum level requirements.
According to some embodiments, a level situation may become even better with future automated test equipment products, wherein the drivers (for example the driver 510) may provide a larger level when compared to that of drivers of conventional automated test equipment products.
In the following, some simple spice simulation results for a lossless case will be described making reference to
A slight difference of the input capacitances of the devices under test 584a to 584d is considered. For example, it is assumed that the first device-under-test 584a has an input capacitance of 2.1 pF, while the other devices-under-test 584b to 584d comprise an input capacitance of 2 pF.
As can be seen from the graphical representation 600, the temporal evolution of the input signal shown by the curve 614 reaches a level of 400 mV, approximately 1 nanosecond after the pulse provided by the driver 510. It can also be seen that after a time T=1.0 nanosecond, a variation of the device-under-test input voltage shown by the curve 614 is relatively small, even in the presence of a small difference of the input capacitances of the devices-under-test.
To summarize the above, in some embodiments according to the invention, for example in the embodiment shown in
In the following, a method for distributing a signal from a driver to a plurality of devices will be described taking reference to
The method 1200 also comprises forwarding 1240 a signal portion incident to the node via the second signal guiding structure, to the first signal guiding structure and to the matching element, while suppressing a reflection of the reflected signal portion incident to the node via the second signal guiding structure back towards the second signal guiding structure. It should be noted that the method 1200 can be supplemented by any of the functionalities described herein.
The Y-sharing topology 1300 is very similar to the Y-sharing topology described with reference
The Y-sharing topology 1300 comprises a driver or buffer 1310 (which is similar to the driver or buffer a 510), a cable 1320 (which is similar to the cable 520), a branch-via or fork via 1340, a resistor 1354 (which is similar to the resistor 554), a second transmission line 1360 (which is similar to the second transmission line 560) and a branching node 1370 (which is comparable to the branching node 570). Moreover, the Y-sharing topology 1300 comprises N branch transmission lines 1380a to 1380n. The N branch transmission lines 1380a to 1380n are circuited between the branching node 1370 and the device connections 1382a to 1382n. The device-under-test connections 1382a to 1382n may be equivalent to the device connections 582a to 582d. Moreover, devices 1384a to 1384n may be connectable, or may be connected, to the device connections 1382a to 1382n.
In the Y-sharing topology 1300, a first end of the branch via 1340 may for example be coupled to the driver or buffer 1310 via the cable 1320, which cable may serve as a first transmission line. The cable or first transmission line 1320 may for example comprise a characteristic impedance ZTL1. A second end of the branch via 1340 may for example be coupled to a first terminal of the resistor 1354. A second terminal of the resistor 1354 may be coupled to a reference potential or grounded potential, or to another fixed potential. A tap 1350 of the branch via or fork via 1340 may be coupled with the branching node 1370 via the second transmission line 1360. The second transmission line 1360 may comprise a characteristic impedance ZTL2. Further, the branch transmission lines 1380a 1380n comprise a characteristic impedance ZTL3.
It has to be noted that in the embodiment shown in
However, different numbers of branches and devices under test can be used as well.
In an embodiment according to invention, the following conditions may be given or fulfilled;
0<ZTL3<ZTL1*N
ZTL2=ZTL3/N; and
Rm=(ZTL1*ZTL2)/(ZTL1−ZTL2).
In a typical embodiment, the characteristic impedance of the first transmission line 1320, which is designated with ZTL1, may be equal to 50 Ohm (ZTL1=50 Ohm). Also, in a typical embodiment, the characteristic impedance of the branch transmission lines 1380a to 1380n, also designated with ZTL3, may lie within a range between 0 and 100 Ohm (0≦ZTL3≦100 Ohm).
However, other ranges for the characteristic impedances may be used in some other embodiments.
Also, in some embodiments the length of the second transmission line 1360 may be short. In some embodiments, the length of the second transmission line 1360 may even be 0. In other words, the second transmission line 1360 may be omitted in some embodiments.
In the following, possible implementation of a fork via structure will be described taking reference to
The structure 1400 comprises a first transmission line 1420, which may be equivalent to the first transmission line 1320. Moreover, the fork via structure 1400 comprises a branch via or a fork via 1440, which may for example be equivalent to the fork via 1340 shown in
However, the structure shown in
However, it should be noted that in the fork via implementation shown in
The propagation delays between the different branches (more precisely, between the fork-via-ends of the branch transmission lines 1480a to 1480d) caused by the “asymmetric” via (or the asymmetric layer structure) may somewhat degrade a performance.
However, the structure shown in
To summarize,
The fork via structures 1600 may further comprise a signal distribution structure 1660. The signal splitting structure 1660 may comprise a plurality of conductive traces 1662a to 1662d. The conductive traces 1662a to 1662d may for example be arranged in a common conductive layer of the multi-layer printed circuit board. The different conductive traces 1662a to 1662d may for example be coupled to the fork via 1650, and may for example extend outwardly from the fork via 1650 into different directions.
However, different geometrical arrangements of the signal splitting structure 1660 may be used. For example, the signal splitting structure 1660 may for example comprise a relatively short common conductor, which is coupled between the fork via 1650 and a branching point, from which branches extend in different directions.
Moreover, the fork via structure 1600 comprises a plurality of branch transmission lines 1680a to 1680d. For example, the branch transmission line 1680a to 1680e may be equivalent to the branch transmission lines 1380a to 1380n. In an embodiment, the signal splitting structure 1660 may be arranged in a layer between the first end of the fork via 1650 and the second end of the fork via 1650. For example, the signal splitting structure 1660 may be arranged in a layer Lm of the multi-layer printed circuit board, which layer Lm is arranged between a layer Ln on which the first transmission line 1620 is formed, and a layer on which the resistor 1654 is arranged. In other words, the signal splitting structure 1660 may be formed on one of the inner layers of the multi-layer printed circuit board.
Furthermore, the conductive traces 1662a to 1662d may be connected to the branch transmission lines 1680a to 1680d using vias 1664a to 1664d. For example, one or more of the branch transmission lines (for example the branch transmission line 1680a, 1680b) may be arranged in a layer of the multi-layer printed circuit board which is on one side (for example above, or below) the layer Lm in which the signal splitting structure 1660 is arranged. Moreover, one or more of the branch transmission line (for example branch transmission lines 1680c, 1680d) may be arranged in one or more layers located on a second side (for example below, or above) the layer Lm, in which the signal splitting structure 1660 is arranged.
Assuming for example that the multi-layer printed circuit board comprises a sequence of conductive layers designated with Lm−2, Lm−1, Lm, Lm+1, Lm+2, in the given order shown in
Similarly, the layer Lm, in which the signal splitting structure 1660 is arranged, may be arranged between the layers Lm−2 and Lm+2, in which the first branch transmission line 1680a and the third branch transmission line 1680c are arranged, as shown in
Accordingly, branch transmission lines are arranged on different sides with respect to the layer LM in which the signal splitting structure 1660 is arranged. Accordingly, propagation delay differences of signals propagating from the first transmission lines 1620 to the different branch transmission lines 1680a to 1680d can be reduced, for example when compared to the structure 1400 shown in
For example in an embodiment there may be only two branch transmission lines, for the example the branch transmission lines 1680a and 1680c. Accordingly, the signal splitting structure 1660 may comprise only two branches. The two branch transmission lines 1680a, 1680c may be coupled with the branch via or fork via 1650 using the signal splitting structure 1660 and the additional vias 1664a, 1664c. In this case, propagation delay be between the first transmission line 1620 and a branch-via-sided end of the branch transmission line 1680a may be equal, for example within a tolerance range of +/−2 picoseconds to a propagation delay between the first transmission line 1620 and a branch-via-sided end of the branch transmission line 1680c. Further, in this case, only the conductive traces 1662a, 1662c of the signal splitting structure 1660 may be present, while the conductive structures 1662b, 1662d may be absent.
Using the above described arrangement, it can be achieved that the branch transmission lines 1680a, 1680c can be arranged on different layers of the multi-layer printed circuit board, while a propagation delay between the first transmission line 1620 and said branch transmission lines 1680a, 1680c is approximately identical.
In another embodiment, there may be actually four branch transmission lines 1680a to 1680d, as shown in
However, using said arrangement the branch transmission lines 1680a to 1680d can be routed on different layers of the multi-layer printed circuit board. A sufficient signal integrity can be maintained, as propagation delay differences between the device-under-test sided ends of the branch transmission lines 1680a to 1680d and a coupling point 1650a, at which the branch transmission line paths split up, are kept small. In other words, using the branch via structure 1600 shown in
To summarize the above, an improved fork via structure or branch via structure 1600 has been described taking reference to
In the following, a short comparisons will be given of the branch via structures 1400 and 1600. As can be seen, the branches or branch transmission lines 1480a to 1480d and 1680a to 1680d are arranged in different layers (of the multi-layered printed circuit board). However, in the branch via structure 1400, the branches are attached to the feeding line (first transmission line 1420) using a via. This structure causes an asymmetry in the signal propagation along the via in a vertical direction, which reduces a reflection cancellation (or which makes the reflection cancellation less effective, or even ineffective in the worst case). Accordingly, the structure 1400 brings along some degradation of signal integrity at some or all of the device-under-test sites. However, the structure 1400 can be used depending on the actual requirements with respect to signal integrity. Nevertheless, an improvement can be obtained using a structure 1600 shown in
To summarize the above,
Moreover, it should be noted that the termination resistor 1654 may also be designated as “fork resistor”.
Also, the first transmission line 1420 may be considered as a feeding line, which may for example guide a signal from a so-called “pin electronic driver” channel module (for example from a channel module of a device tester) towards the branch via or fork via 1650.
As can be seen from
In the following, some further explanations will be given taking reference to
When the reflections from both branch ends arrive again at the fork point 1810, a portion will be reflected back to the branch end again and another portion will be reflected into the feeding line and into the other branch end. If now the reflected portion from one branch end the refracted portion from the other branch end cancel out each other, this type of Y-sharing will work properly up to the highest speeds without any signal distortion. To achieve this, theoretical situation, a perfect symmetry between both branches 1814, 1816 may be used (for example with respect to a trace length and an impedance or input impedance of the device-under-test). Furthermore, a certain impedance ratio between the feeding line 1804 and the branch lines 1814, 1816 may be used to fulfill the reflection cancellation condition. These impedances can be calculated from transmission line theory.
The reflection coefficient r for the signal that propagates from the branch end to the fork point. 18 and the refraction coefficient b for the signal that is refracted into the other branch end are given by:
Thus, if it is desired that the reflected and the refracted portion cancel out each other, the requirement for the impedance ratio is Z1/Z2=2. For a 50Ω feeding line 1804 from the tester (or from output driver or output buffer 1802 of the tester) this means that the branch lines 1814, 1816 need to have an impedance of 100Ω. Interestingly, this also the matching condition for the signal that approaches the fork point. 1810 from the feeding line 1804, so that no energy is lost when driving into the fork 1410 for example from the feeding line 1804. An advantage of the Y-sharing is the symmetry. The symmetry ensures that (in an ideal case) all devices under test (DUTs) see the same signal. So, for example all device inputs are charged with the same signal rise time, which is not the case for the Daisy-Chain sharing. Furthermore, devices that have unequal input impedances on different pins (e.g., stacked die devices) can be tested easily with Y-sharing, since the propagation delay from the feeding point to the input pins of the shared pins are identical by design and the different input signals can be calibrated individually. This is not the case for Daisy-Chain sharing. Therefore, stacked die testing is not possible with the Daisy-Chain sharing approach, but with Y-sharing.
To summarize the above, under the conditions described for the impedances, a canceling of reflected and refracted signal portions can be obtained in the circuit shown in
Theoretically, the simple conventional type of Y-sharing can also be expanded to a fan-out factor of 4. However, this idea can hardly be implemented in a realistic printed-circuit-board (PCB) process.
In the following, some further embodiments according to the invention will be described. However, it should be noted that in some of the embodiments described in the following, a cancellation of reflected and refracted signal portions will also be exploited.
The circuit 2000 further comprises an optional second transmission lines 2060, which may comprise an impedance of Z2, and which second transmission line may be equivalent to the second transmission line 1360. The second transmission line 2060 is circuited between the node 2050 and a branch node or fork node 2070, which may for example be equivalent to the branch node or fork node 1370. However, in the absence of the second transmission line 2060, the node 2050 may coincide with the branch node or fork node 2070.
The circuit 2000 further comprises a plurality of N branch transmission line 2080a to 2080n, which branch transmission lines 2080a to 2080n may branch from the branch node or fork node 2070. Furthermore, the circuit 2000 may for example comprise N device-under-test connections 2082a to 2082n, which may for example be equivalent to the device-under-test connections 1382a to 1382n. Further, there may be a possibility to connect N devices-under-test 2084a to 2084n to the device-under-test connections 2082a to 2082n. For example, each of the branch transmission lines 2080a to 2080n may be associated with one device-under-test connection 2082a to 2082n, or with one device-under-test 2084a to 2084n. Thus, each of the branch transmission lines 2080a to 2080n may connect one of the device-under-test connections 2082a to 2082n with the branching node or fork node 2070. However, in some other embodiments, more than one device under test connection may be coupled to a single branching line.
The so-called new “laqi-b” approach for a by-N Y-sharing uses, at least partially, similar principles or even the same principles as the conventional approach to avoid reflections. This means, it is advantageous to design the branches absolutely symmetrical. Also, it is desired that the impedance ratio Z3/Z2 between the N branches and the feeding line may be chosen such that the reflected signal portions and the refracted signal portions cancel out each other (for example as described with reference to
However, it is the key idea of some embodiments of the present invention to add a so-called “fork resistor” having a resistance value R (for example, the resistor 2054), such that the useful trace impedances can be shifted into an impedance range, which is producible (or even producible with moderate effort) with standard printed circuit board processes.
The values for the so-called port resistor (resistor 2054) and the traced impedances may be chosen in the following way:
A desired characteristic impedance Z3 of the branch transmission line 2080a to 2080n may be given with
0<Z3<Z1*N.
Consequently, the impedance Z2 of the second transmission line 2060 and the resistance R of the fork resistor 2054 may be chosen according to the following equation:
Z
2
=Z
3
/N; and
R=(Z1*Z2)/(Z1−Z2) .
A length L of the second transmission line 2060 may be chosen arbitrarily. In a special case, the length L may reach a value of 0, which means that the second transmission line 2060 can be omitted.
It should be noted here that naturally the impedance Z2 of the second transmission line 2060 and the resistance R of the fork resistor 2054 may deviate from the ideal values defined by the above equations in accordance with acceptable tolerances. For example, a tolerance of +/−20% from the ideal desired values may be acceptable in some applications. In other applications, a maximum tolerance of, for example, +/−10% or +/−−5%, may be desirable.
Moreover, it should be noted that the value R of the resistance increases if the impedance value Z3/N approaches the value of Z1. However, in practical applications, it is typically desirable that Z3/N differs from the impedance Z1 by at least 20% or even by at least 50%. Accordingly, the resistance of the resistor 2054 is smaller than ten times the impedance Z. In many cases, the resistance R of the resistor 2054 is even smaller than the characteristic impedance Z1.
In the following, some further embodiments will be described taking reference to
To avoid the reduced swing disadvantage, the following setting can be used for N=4, Z1=50Ω, Z3=100Ω, Z2=25Ω and R=500. The result is a swing identical to the by-4 Daisy-Chain sharing (½ of the programmed driver swing) and slightly increased rise times, since in this case, the device-under-test input capacitances are charged from a source impedance of 100Ω. Again, the trace segment (second transmission line) with Z2=25Ωcan be omitted. However, a 100Ω trace impedance can still be produced reasonably in a state-of-the-art printed circuit board process.
However, it should be noted that the impedance of the branch transmission lines 2080a to 2080n could be varied in accordance with the requirements. For example, branch impedances between 50Ω and 100Ω can be fabricated in a technologically advantageous way. However, in some printed circuit board processes, it is difficult to obtain transmission lines having an impedance as high as 100Ω. In such processes, it may be advantageous sometimes to use branch transmission lines having a characteristic impedance between 60Ω and 80Ω. Moreover, it should be noted that in some embodiments, it is desired to have a relatively high impedance of the branch transmission lines to obtain a large voltage swing at the device-under-test connections 2082a to 2082n. On the other hand, it is sometimes desired to keep the characteristic impedances of the branch transmission lines as low as possible in order to obtain short rise times of the edges arising at the device-under-test connections 2080a to 2080n. Thus, in some embodiments, the characteristic impedance of the branch transmission lines 2080a to 2080n will be chosen to obtain a trade-off between manufacturability, swing and rise times.
It should be noted here that a nominal impedance or desired impedance of the fork resistor 2054 depends on the characteristic impedance of the branch transmission lines, as described above.
In the following, simulation results will be described.
As can be seen from a comparison of
For the conventional Y-sharing limited to a fan-out of 2 as well as for the laqi-b sharing, it is desired to design the branches absolutely symmetrical (or at least approximately symmetrical) to exploit the reflection cancellation effect. Due to manufacturing limitations for the printed circuit board and input capacitance variations between the devices under test however, the theoretical symmetry (or the desired symmetry) not can be achieved completely. Therefore, the reflections will not be cancelled completely, resulting in remaining signal distortions.
A means to further reduce this effect is to introduce a complete or incomplete termination at the branch ends to reduce the initial reflections at the devices under test. The fact that the device-under-test input capacitance acts like a short circuit in the first instance of time, however, prevents a complete matching at the branch end. Therefore, the reflection cancellation effect at the fork point still remains important, and the requirement for a well chosen trace impedance ratio and the fork resistance for the laqi-b version of Y-sharing stays in place. Nevertheless, this type of termination not only improves the signal integrity, but also improves the rise time. The penalty, however, is that it reduces the swing again depending on what value is used for termination. A completely matched termination would reduce the swing to 1/Nth of the programmed driver level.
As mentioned above, the termination resistors 2790a to 2790n will cause a termination of the branch transmission lines and, therefore, increase a matching. Accordingly, reflections at a test socket for a device under test or at an input of the device under test can be reduced. The resistance RT may, for example, be chosen to be larger than or equal to the characteristic impedance Z3 of the branch transmission lines.
Embodiments according to the invention may for example be applied in high speed memory testing DDR2 devices. In some embodiments, data rates up to 1033 Mbps can be achieved. However, in further embodiments, even higher data rates may be achieved.
Some embodiments according to the invention can be applied for a multi-site testing. For example, a multi-site testing×64 may be performed. However, embodiments according to the invention can also be applied in a multi-site testing having lower or even higher sharing factors. In some embodiments, a plurality of socket boards (for example 16 socket boards) can be used, each of the socket boards providing a device-under-test socket for two or more devices (for example for two or four devices).
Some embodiments according to the invention can be applied in a multi-site testing×128. For example, 32 socket boards may be used in combination with a by-4 sharing. The multi site testing may run up to 2.5 Gbps. A new laqi-b sharing concept may contribute in achieving these goals.
The test adapter 2900 may for example be applied as a complete DDR2 interface for multi-site testing×64 using a laqi-sharing with a fan-out-factor of 2 or with a fan-out-factor of 4.
In some systems, the case N=2 is an advantageous embodiment. In some other systems, the case N=4 is an advantageous embodiment. However, other values for N can be used, in dependence on the specific requirements.
In some embodiments, the branch point 214 is a via designed with high diligence or accurateness to have good symmetry. Otherwise (in the absence of good symmetry), there may be signal distortions, which may be tolerable in some cases, and which may need to be avoided in some other cases.
While this invention has been described in terms of several embodiments, there are alterations, permutations, and equivalents which fall within the scope of this invention. It should also be noted that there are many alternative ways of implementing the methods and compositions of the present invention. It is therefore intended that the following appended claims be interpreted as including all such alterations, permutations and equivalents as fall within the true spirit and scope of the present invention.
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/EP2008/007913 | 9/19/2008 | WO | 00 | 4/26/2011 |
